TWI406187B - Fast and high quality image/video interpolation method and apparatus - Google Patents

Abstract

The invention provides a fast and high quality image/video interpolation method and apparatus. An up-sampling step first estimates the diagonal missing pixels and interpolates the remaining missing pixels based on the estimation result. An edge-sensing interpolator is provided to guide the interpolation automatically along the edges, where local gradients are non-linearly mapped to achieve the weights. A post-processing refinement step then updates the interpolated high resolution image iteratively, by adding the error between the observed low resolution image and the simulated low resolution image, to attain a clear, natural and visually pleasuring high resolution output.

Description

Fast high-definition video image interpolation method and device

The invention relates to the technical field of image processing, in particular to a method and a device for interpolating fast high-definition video images.

With the rapid development of display panel technology, large-scale ultra-high resolution (Quad Full HD, ie 3841×2160 resolution) multimedia flat-panel TVs and players will become one of the key electronic product projects to be developed in the future. In order to enable existing high-resolution image and video media to be played on these ultra-high-resolution display panels, the development of fast high-quality image interpolation methods is a future development trend.

In the US Patent No. 5,001,563, the adjacent image to be interpolated is filtered to generate an interpolation point, which focuses on the design of the filter. However, the use of the filter method only refers to the adjacent image of the interpolation point, and the image that can be interpolated for the image information that needs to be far away, such as the image with low inclination and thin oblique line, because of insufficient image information, the interpolated value The image will be jagged and zigzag due to misjudgment.

In order to compensate for the problem of image misjudgment and ambiguity caused by insufficient image information, the US Patent No. 6,133,957 proposes the use of the Least texture variance method, which is used to find the texture in the image to be interpolated. The smallest change is used as the source of the interpolated image. At the same time, in order to compensate for the lack of image misjudgment and the lack of blur due to insufficient image information, the minimum rule of texture change is to find the source of the interpolated image by finding the smallest variation in the image. However, due to this law, there are often some confusing situations, such as the occurrence of more than one trace change, etc. In this case, the probability of image misjudgment is also greatly increased, resulting in unnatural images. In order to avoid misjudgment, it adds a mechanism of Confidence judgment, which uses the complexity of the texture change to give the degree of credibility. When the credibility is small, the image source is interpolated in the vertical direction. To avoid possible misjudgments. However, this method causes image discontinuity due to the fact that most of the images are interpolated in the vertical direction.

It can be seen from the foregoing that the conventional image interpolation method only focuses on the pursuit of image quality, and often neglects the consideration of the amount of calculation. In addition, the display panel technology has been rapidly developed in recent years, and the resolution of the panel is getting higher and higher, which makes the conventional high computational amount. The feasibility of interpolation in real-time applications is getting lower and lower. Therefore, there are many shortcomings in the conventional interpolation technique for increasing image resolution, and it is necessary to improve it.

The main object of the present invention is to provide a fast high-definition video image interpolation method, which can calculate the interpolation value of the blank pixel according to the texture of the image pattern by weighted average interpolation method, and repeatedly utilize the original input low resolution. The error between the image and the low-resolution image simulated by the high-resolution image after interpolation, the high-resolution image is corrected to make the output image clearer, more natural and true.

According to a feature of the present invention, the present invention provides a fast high-definition video image interpolation method for use in a single frequency multiplier (SCALER) to insert pixels in a low-resolution image to increase the low-resolution image. The resolution, 俾 produces a high-resolution image consisting of a plurality of pixels arranged in a two-dimensional array, including pixels X _{2 i , 2 j} , X _{2 i , 2 j +2} , _{X 2 i} +2,2 _{j +2} and _{X 2 i +2,2 j, i,} j is a positive integer, the image interpolation method comprising the steps of: (a) calculating a diagonal blank pixels X _{2 i +1 , 2 j +1} the absolute pixel difference of its neighboring upper left, upper right, lower right, and lower left directions; (B) calculating the oblique according to the absolute pixel difference in step (A) and a nonlinear function The _weights to the upper left, upper right, lower right, and lower left directions of the blank pixels X _{2 i +1} , _{2 j +1} ; (C) calculate the oblique blank pixels according to the weights in step (B) (D) calculating the absolute pixel difference of a remaining blank pixel X _{2 i , 2 j +1 in} the direction of the known pixel adjacent to the four corners; (E) according to the absolute pixel difference in step (D) and the non-line Of the function to calculate the remaining blank pixels X _{2 i,} up right _{2 j +1} heavy, the weight W _D downward, leftward, the weight W _L and the right weight W _R; (F) calculates the remaining blank pixels X _{2 i, 2 j +1} horizontal absolute pixel difference (Δ H ) and vertical absolute pixel difference (Δ V ); (G) absolute pixel difference (Δ H ) in the horizontal direction and absolute pixel difference in the vertical direction in step (F) (Δ V ) to calculate the value of the remaining blank pixels X _{2 i , 2 j +1} .

According to another feature of the present invention, the present invention provides a fast high-definition video image interpolation device that receives a low-resolution image and inserts pixels into the low-resolution image to increase the resolution thereof, thereby generating a high resolution. a low-resolution image consisting of a plurality of pixels arranged in a two-dimensional array, wherein the pixels X _{2 i , 2 j} , i , j are positive integers, and the image interpolation device includes a video image interpolation A processor, a line buffer, a table device, and an image enhancement post processor. The interpolation within video image processor receives the low-resolution image, and using a non-linear exponential function (exponential function), and a blank pixel X on the obliquely adjacent left, upper right _{2 i +1,2 j +1,} the left and right pixel absolute difference of the four directions, right _{+1,2 j + 1} is calculated and the values of the neighboring pixels X _{2 I} oblique blank pixels, and weights based on the diagonal of the blank and the pixel X _{2 i +1,2 j +1} value adjacent to the pixel, the oblique blank pixels X _{2 i +1,2 j +1} as a weighted average interpolation. The line buffer is coupled to the video image interpolating processor to temporarily store the data generated by the video image interpolating processor. The table device is connected to the video image interpolating processor, and uses the look-up table method to perform an instant operation on the exponential function according to the video image interpolating processor output index. The image enhancement post processor is coupled to the line buffer device for image post-processing of the high resolution image.

The invention provides a fast video image interpolation method. The invention firstly uses a pixel lifting technique to amplify the input low-resolution image, and then uses an image enhancement post-processing technique to repeatedly correct the amplified high-resolution image.

1 is a block diagram of a fast high definition video image interpolation device of the present invention. The interpolation device 100 receives a low-resolution image I _L and inserts pixels into the low-resolution image to increase the resolution thereof, and generates a high-resolution image I _H (0), and the low-resolution image I _L 2A is a plurality of pixels X _0,0 , X _0,2 , X _0,4 ... ; X _2,0 , X _2,2 , X _2,4 . . arranged in a two-dimensional array. _{.; X 4,0, X 4,2,} X 4,4 ... , where the contents including the pixel _{X 2 i, 2 j, X} 2 i, 2 j +2, X 2 i +2,2 j + ₂ and X _{2 i +2} , _{2 j} , i , j are positive integers, and the high-resolution image I _H (0) generated by the inserted pixels is as shown in FIG. 2B, and becomes a majority arranged by a two-dimensional array. Pixels X _0,0 , X _0,1 , X _0,2 ... ;X _1,0 , X _1,1 , X _1,2 ... ;X _2,0 , X _2,1 , X _{2 , 2} ... ; X _3,0 , X _3,1 , X _3,2 ... ;X _4,0 , X _4,1 , X _4,2 ..., the image interpolation device includes A video image interpolating processor 110, a line buffer 120, a table device 130, and an image enhancement post processor 140.

The interpolation within video image processor 110 receives the low-resolution image I _L, and the use of a non-linear exponential function (exponential function), and a diagonal blank pixels X _{2 i +1,2 j +1} adjacent to the left, upper right, lower right and lower left of absolute pixel differences of the four directions calculated _{2 i +1,2 j +1} and right adjacent pixel value to the slant of blank pixels X, and based on the weight and slant of the _{2 i +1,2 j +1} pixel values of neighboring pixels of the blank X, the oblique blank pixels X _{2 i +1,2 j +1} as a weighted average interpolation.

The line buffer 120 is coupled to the video image interpolating processor 110 to temporarily store the data generated by the video image interpolating processor 110. The table device 130 is connected to the video image interpolating processor 110, and performs an instant operation on the exponential function by using a look-up table according to the index output by the video image interpolating processor 110. The image enhancement processor 140 is coupled to the line buffer device 120 to perform image enhancement post processing on the high resolution image I _H (0).

The pixel lifting technique used in the present invention uses a nonlinear exponential function and an absolute difference between adjacent pixels to calculate the weight of a blank pixel to a known pixel of the neighboring square, and according to the weight and its neighboring Known pixel values, weighted average interpolation of the blank pixels. In order to make the interpolation process more efficient, the interpolation method first interpolates the oblique blank pixels, and then interpolates the remaining blank pixels according to the interpolation result and the pixels of the original image.

2 is a flow chart of a fast high quality video image interpolation method of the present invention. The image interpolation method is used in a single frequency multiplier (SCALER) to insert a pixel into a low resolution image I _L to increase the resolution of the low resolution image, and generate a high resolution image I _H ( 0), the low-resolution image I _L lines arranged in a two-dimensional array consisting of a plurality of pixels, wherein the pixel comprises _{X 2 i, 2 j, X} 2 i, 2 j +2, X 2 i +2,2 _{j +2} and _{X 2 i +2,2 j, i,} j is a positive integer.

Referring also to FIG. 4 The present invention is a high resolution image I _H (0) in a diagonal blank pixels X _{2 i +1,2 j +1} Schematic known neighboring pixels. In this interpolation, the oblique blank pixels X _{2 i +1,2 j +1} adjacent the four corners of the known pixel interpolation, _{X 2 i, 2 j, X} 2 i, 2 j +2, X 2 _{i +2,2 j +2} and X _{2 i + 2,2 j are} obtained after weighted average interpolation, wherein the pixels X _{2 i , 2 j} , X _{2 i , 2 j +2} , X _{2 i +2, 2 j +2} and _{X 2 i} +2,2 _j-based low-resolution image I _L is known in the pixel.

First, in step (A), the absolute pixel difference of the adjacent pixel direction of the blank pixel X _{2 i + 1, 2 j +1 is} calculated, that is, the upper left absolute pixel difference D _UL and the upper right absolute pixel difference D _UR , the absolute pixel difference D _DR to the lower right and the absolute pixel difference D _DL to the left. The formulas are:

D _UL =| X _{2 i ,2 j} - X _{2 i +2,2 j +2} |+| X _{2 i ,2 j -2} - X _{2 i +2,2 j} |+| X _{2 i -2, 2 j} - X _{2 i , 2 j +2} |,

D _UR =| X _{2 i ,2 j +2} - X _{2 i +2,2 j} |+| X _{2 i -2,2 j +2} - X _{2 i ,2 j} |+| X _{2 i ,2 j +4} - X _{2 i +2,2 j +2} |,

_{D DR = | X 2 i,} 2 j - X 2 i +2,2 j +2 | + | X 2 i +2,2 j - X 2 i +4,2 j +2 | + | X 2 i, _{2 j +2} - X _{2 i +2,2 j +4} |,

D _DL =| X _{2 i ,2 j +2} - X _{2 i +2,2 j} |+| X _{2 i ,2 j} - X _{2 i +2,2 j -2} |+| X _{2 i +2, 2 j +2} - X _{2 i +4,2 j} |.

In step (B), a step based on the absolute pixel difference (A), and a nonlinear function to calculate a diagonal blank pixels X _{2 i +1,2 j +1} upper left, upper right, lower right, and The weight to the left four directions. Wherein, the nonlinear function in the step (B) is an exponential function.

That is, according to calculation step (A) of the results, by using an exponential function, calculates weights of the white pixel X direction pixel adjacent corners of known weight _{2 i +1,2 j +1,} i.e., to the left the weighting W _UL, rightward The weight W _UR , the right lower weight W _DR and the lower left weight W _DL . The formulas are

W _UL =exp(- D _UL / h ),

W _UR =exp(- D _UR / h ),

W _DR =exp(- D _DR / h ),

W _DL =exp(- D _DL / h ),

Where h is an adjustment parameter, 0 < h , in the preferred embodiment, the value of h is 48.

In step (C), the value of the oblique blank pixel is calculated according to the weight in step (B). The interpolated pixel X oblique blank _{2 i +1,2 j +1} is:

It can be seen from the above steps that the weight and the pixel difference are nonlinearly inversely proportional, so that the adjacent pixel will become larger due to the absolute pixel difference of its corresponding direction, and the proportion of its own participation in the interpolation is reduced, and vice versa. .

Therefore, the interpolation method of the present invention can automatically interpolate the oblique blank pixels along the edge of the image, so that the image texture of the image can remain clear and complete after the image is enlarged. In addition, because of the use of the nonlinear inverse ratio function for weight calculation, the high resolution image output by the interpolation method of the present invention is clearer than the high resolution image output by the conventional weighted average interpolation method using the linear inverse ratio function. sharp.

According to the foregoing steps (A) to (C), after the interpolation operation is performed on all the oblique blank pixels, the interpolation method of the present invention then interpolates the remaining blank pixels. FIG. 5 is a schematic diagram of each of the remaining blank pixels X _{2 i , 2 j +1} and its neighboring known pixels in a high-resolution image I _H (0) of the present invention. As shown in the high-resolution image I _H (0) through steps (A) to step (C), the sum of all the diagonal-blank pixel interpolation operation performed X _{2 i +1,2 j +1} 5 The arrangement of the pixels. As shown in FIG. 5, each of the remaining blank pixels X _{2 i , 2 j +1} has been surrounded by four adjacent known pixels from four directions of up, down, left, and right. In order to avoid the zipper effect after the image is enlarged, the internal difference method of the present invention first determines the remaining blank pixels X _{2 i , 2 j +1} according to the absolute pixel difference between the horizontal and vertical directions of the adjacent area. The interpolation direction is then interpolated by the weighted average interpolation method along the determined direction to the remaining blank pixels X _{2 i , 2 j +1} .

In step (D), the absolute pixel difference of the remaining blank pixels X _{2 i , 2 j +1 in} the direction of the known pixel adjacent to the four corners is calculated. Wherein, in the step (D), the following formula is used to calculate the adjacent absolute pixel difference D _U , the downward absolute pixel difference D _D , and the absolute left pixel difference of the remaining blank pixels X _{2 i , 2 j +1} D _L and absolute right pixel difference D _R :

D _U =| X _{2 i -1,2 j +1} - X _{2 i +1,2 j +1} |+| X _{2 i -2,2 j} - X _{2 i ,2 j} |+| X _{2 i - 2,2 j +2} - X _{2 i ,2 j +2} |,

D _D =| X _{2 i -1,2 j +1} - X _{2 i +1,2 j + 1} |+| X _{2 i ,2 j} - X _{2 i +2,2 j} |+| X _{2 i , 2 j +2} - X _{2 i +2,2 j +2} |,

D _L =| X _{2 i ,2 j} - X _{2 i ,2 j +2} |+| X _{2 i -1,2 j -1} - X _{2 i -1,2 j +1} |+| X _{2 i + 1,2 j -1} - X _{2 i +1,2 j +1} |,

D _R =| X _{2 i ,2 j} - X _{2 i ,2 j +2} |+| X _{2 i -1,2 j +1} - X _{2 i -1,2 j +3} |+| X _{2 i + 1,2 j +1} - X _{2 i +1,2 j +3} |,

_{Among, X 2 i - 1,2 j +1} , X 2 i +1,2 j -1, X 2 i -1,2 j -1, X 2 i +1,2 j +1, X 2 i - _{1,2 j +3, X 2 i +1,2} j +3 lines the step (C), the calculated value of the diagonal of the blank pixels.

In step (E), the absolute weight W _U , the downward weight W _D , and the left weight are calculated according to the absolute pixel difference in the step (D) and the nonlinear function to calculate the remaining blank pixels X _{2 i , 2 j +1} W _L and rightward weight W _R . The formulas are:

W _U =exp(- D _U / h ),

W _D =exp(- D _D / h ),

W _L =exp(- D _L / h ),

W _R =exp(- D _R / h ),

Where h is the aforementioned adjustment parameter.

In step (F), the calculation of the remaining blank pixels X _{2 i, 2 j +1} in the horizontal direction of the absolute pixel differences (Δ H) and the vertical direction of the absolute pixel differences (Δ V). Wherein the step (F) in the system using the following formula to calculate the remaining blank pixels X _{2 i, 2 j +1} in the horizontal direction of the absolute pixel differences (Δ H) and the vertical direction of the absolute pixel differences (Δ V):

In the step (G), the values of the remaining blank pixels X _{2 i , 2 j + 1} are calculated according to the horizontal absolute pixel difference (Δ H ) and the vertical absolute pixel difference (Δ V ) in the step (F). Wherein the step (G) in Comparative horizontal line direction of the absolute pixel differences (Δ H) and the vertical direction of the absolute pixel differences (Δ V) of the value, determining the remaining blank pixels X _{2 i, 2 j +1} interpolation direction and the Its interpolation value.

When the horizontal direction of the absolute pixel differences (Δ H) is greater than the vertical direction of the absolute pixel differences (Δ V), the remaining blank pixels X _{2 i, 2 j +1} in the vertical direction for interpolation, which is interpolated:

When the horizontal direction of the absolute pixel differences (Δ H) is smaller than the vertical direction of the absolute pixel differences (Δ V), the remaining blank pixels X _{2 i, 2 j + 1} in the horizontal direction for interpolation, which is interpolated:

When the horizontal direction absolute pixel differences (Δ H) is equal to absolute vertical pixel differences (Δ V), the remaining blank pixels X _{2 i, 2 j +1} within the interpolation is:

By repeating this step (A) to step (G), the high-resolution image I _H (0) can be generated. In order to achieve the low computational complexity in practical applications, the present invention avoids performing an instant operation on the exponential function, and instead uses a look-up table method to obtain the weight values required for interpolation.

Taking an 8-bit image as an example, from the above steps (A) and (D), it can be known that the effective range of each absolute pixel difference is 0-765, so the present invention only needs to utilize one The table device 130 of the 766 field makes a query, so that the required weight can be obtained instantly without the need of an amount of calculation, wherein the operation result of each field in the table device 130 is exp( -k / h ). , k is 0, 1, 2, ..., 765.

In addition, if the input image is a 16-bit (16-bit) image, the present invention first divides the absolute pixel difference obtained in steps (A) and (D) by 256, and rounds the result to an integer. The integer is then used as an index to find the desired weight value from the table device 130 for interpolation.

After all the remaining blank pixels are interpolated according to the above steps (D) to (G), all the pixels in the high-resolution image I _H (0) have been obtained, and the entire pixel lifting process is completed.

Next, the present invention performs step (H) to perform a post-enhancement processing on the enlarged high-resolution image, and the enhanced post-processing technique of the present invention repeatedly corrects to enhance the sharpness and image quality of the output image.

The enhanced post-processing technique of the present invention is an iterative image correction technique, which repeatedly uses the error between the low-resolution image simulated by the amplified high-resolution image and the original input low-resolution image, and the high resolution The image is corrected.

Fig. 6 is a schematic view showing the operation principle of the image enhancement post-processing technique used in the present invention. For convenience of description, the present invention uses I _{L to} represent the low-resolution image of the original input, and I _H ( n ) represents the high-resolution image output after the nth correction, where n =0, 1, 2, ..., N , N is the total number of iterations. In the method, in order to reduce the amount of calculation, the preset value of N is 5. Figure 7 is a detailed flow diagram of step (H) of the present invention.

In the step (H1), an initial high-resolution image I _H (0) is set, wherein I _H (0) is the high-resolution image outputted from the foregoing steps (A) to (G).

In step (H2), an error e between the low-resolution image I _L and a simulated low-resolution image simulated by the high-resolution image I _H ( n ) is calculated, where I _H ( n ) is High resolution image at the nth iteration.

Wherein, step (H2) uses the following formula to calculate the error e :

e = I _L -( f * I _H ( n ))↓2,

Among them, (‧) ↓ 2 is a down-sampling program, and f is a 2×2 averaging filter.

In step (H3), the high-resolution image I _H ( n ) is corrected according to the error e , and a corrected high-resolution image I _H ( n +1) is generated. Wherein, the step (H3) uses the following formula to correct the high-resolution image I _H ( n ) and generate the corrected high-resolution image I _H ( n +1):

I _H ( n +1)= I _H ( n )+ g *( e )↑2,

Among them, the operation function ↑2 is a zero order up-sampling program, and g is a 5×5 Gaussian filter whose values are:

In the step (H4), the step (H1) to the step (H3) are repeated until a predetermined number of operations. In this embodiment, the predetermined number of operations is five.

With the completion of the enhanced post-processing steps of the present invention, the entire fast high-definition video image interpolation process is also completed.

It can be seen from the foregoing description that the present invention discloses a fast video image interpolation method, which includes a pixel lifting technique and an image enhancement post processing technique. The pixel lifting technique first interpolates the oblique blank pixels, and then interpolates the remaining blank pixels according to the interpolation result and the pixel values of the original image. It uses a nonlinear function and the absolute difference between adjacent pixels to calculate the weight of a known pixel adjacent to a blank pixel, and uses the weighted average interpolation method according to the weight and its neighboring known pixel values. The texture of the image pattern calculates the interpolation value of the blank pixel. The image enhancement post-processing technique repeatedly uses the error between the original input low-resolution image and the low-resolution image simulated by the interpolated high-resolution image to correct the high-resolution image to make the output image clearer and more natural. And like truth.

In order to achieve the requirements of high image quality and low computational complexity in practical applications, the interpolation method of the present invention changes its traditional purely interpolated design mode, and the concept of post-interpolation and post-enhancement is taken as an outline, and under this outline, for each The technical part is aimed at simple and efficient design. Therefore, compared with the existing interpolation method, the interpolation method not only provides instant image enlargement, but also provides a clear and natural image output, achieving high image quality and low computational complexity.

From the above, it can be seen that the present invention is extremely useful in terms of its purpose, means, and efficacy, both of which are different from those of the prior art. It should be noted that the various embodiments described above are merely illustrative for ease of explanation, and the scope of the invention is intended to be limited by the scope of the claims.

100. . . Interpolation device

110. . . Video image interpolation processor

120. . . Line buffer

130. . . Table device

140. . . Image enhancement processor

Step (A) ~ Step (H)

Step (H1)~Step (H4)

1 is a block diagram of a fast high definition video image interpolation device of the present invention.

2A is a schematic diagram of a low resolution image of the present invention.

2B is a schematic diagram of a high resolution image of the present invention.

3 is a flow chart of a fast high quality video image interpolation method of the present invention.

4 is a schematic diagram of an oblique blank pixel and its adjacent known pixels in a high resolution image of the present invention.

FIG. 5 is a schematic diagram of a remaining blank pixel and its neighboring known pixels in a high resolution image of the present invention.

Fig. 6 is a schematic view showing the operation principle of the image enhancement post-processing technique used in the present invention.

FIG. 7 is a detailed flowchart of image post-enhancement processing of the fast high-definition video image interpolation method of the present invention.

100. . . Interpolation device

110. . . Video image interpolation processor

120. . . Line buffer

130. . . Table device

140. . . Image enhancement processor

Claims (12)

A fast high-definition video image interpolation method used in a single frequency multiplier (SCALER) to insert pixels in a low-resolution image to increase the resolution of the low-resolution image, thereby generating a high resolution image, the low-resolution image of a plurality of pixel lines arranged in the two-dimensional array, where the pixels comprising _{X 2 i, 2 j, X} 2 i, 2 j +2, X 2 i +2,2 j +2 and _{X 2 i +2,2 j, i,} j is a positive integer, the image interpolation method comprising the steps of: (a) calculated on a diagonal blank pixels X _{2 i +1,2 j +1} adjacent leftward Absolute pixel difference in the four directions of the upper right, the lower right, and the lower left; (B) calculating the oblique blank pixel X _{2 i +1, 2} according to the absolute pixel difference in step (A) and a nonlinear function _{j +1} the weights of the upper left, upper right, lower right, and lower left directions; (C) calculate the value of the oblique blank pixel according to the weight in step (B); (D) calculate a remaining blank The pixels X _{2 i , 2 j +1} are absolute pixel differences of the known pixel directions adjacent to the four corners; (E) calculating the remaining blank pixels X according to the absolute pixel difference in the step (D) and the nonlinear function _{2 i , 2 j +1} up weight W _U , down weight W _D , left weight W _L and right weight W _R ; (F) calculating the horizontal direction of the remaining blank pixels X _{2 i , 2 j +1} Absolute pixel difference (Δ H ) and vertical absolute pixel difference (Δ V ); and (G) calculated according to the horizontal absolute pixel difference (Δ H ) and the vertical absolute pixel difference (Δ V ) in step (F) The value of the remaining blank pixels X _{2 i , 2 j +1} . The fast high-definition video image interpolation method according to claim 1, further comprising the step of: (H) performing image enhancement post-processing on the high-resolution image. The fast high-definition video image interpolation method described in claim 2, wherein the steps (A) to (H) are repeatedly performed. The interpolation fast high quality video images of the item 3 of patent application range, wherein the step (A) is based using the following formula to calculate the diagonal blank pixels X _{2 i +1,2 j +1} Its adjacent left upper absolute pixel difference D _UL , upper right absolute pixel difference D _UR , right lower absolute pixel difference D _DR and lower left absolute pixel difference D _DL : D _UL =| X _{2 i , 2 j} -X _{2 i +2,2 j +2} |+| X _{2 i ,2 j -2} - X _{2 i +2,2 j} |+| X _{2 i -2,2 j} - X _{2 i ,2 j +2} |, D _UR =| X _{2 i ,2 j +2} - X _{2 i +2,2 j} |+| X _{2 i -2,2 j +2} - X _{2 i ,2 j} |+| X _{2 i ,2 j + 4} - X _{2 i +2,2 j +2} |, D _DR =| X _{2 i ,2 j} - X _{2 i +2,2 j +2} |+| X _{2 i +2,2 j} - X _{2 i +4,2 j +2} |+| X _{2 i ,2 j +2} - X _{2 i +2,2 j +4} |, D _DL =| X _{2 i ,2 j +2} - X _{2 i +2, 2 j} |+| X _{2 i ,2 j} - X _{2 i +2,2 j -2} |+| X _{2 i +2,2 j +2} - X _{2 i +4,2 j} |. The fast high-definition video image interpolation method according to claim 4, wherein the nonlinear function in the step (B) is an exponential function, that is, the oblique blank pixel X _{2 i +1, 2} The calculation of the left upper weight W _UL , the upper right weight W _UR , the right lower weight W _DR and the lower left weight W _DL of _{j +1} are respectively: W _UL =exp(- D _UL / h ), W _UR =exp( - D _UR / h ), W _DR =exp(- D _DR / h ), W _DL =exp(- D _DL / h ), where h is an adjustment parameter, 0 < h . The fast high-definition video image interpolation method described in claim 5, wherein the following formula is used in the step (C) to calculate the value of the oblique blank pixel X _{2 i +1, 2 j +1} : The fast high-definition video image interpolation method according to claim 6, wherein the step (D) uses the following formula to calculate the remaining blank pixels X _{2 i , 2 j +1} adjacent thereto Upward absolute pixel difference D _U , downward absolute pixel difference D _D , left absolute pixel difference D _L and rightward absolute pixel difference D _R : D _U =| X _{2 i - 1,2 j +1} - X _{2 i + 1,2 j +1} |+| X _{2 i - 2,2 j} - X _{2 i ,2 j} |+| X _{2 i -2,2 j +2} - X _{2 i ,2 j +2} |, D _D =| X _{2 i -1,2 j +1} - X _{2 i + 1,2 j +1} |+| X _{2 i ,2 j} - X _{2 i +2,2 j} |+| X _{2 i ,2 j +2} - X _{2 i +2,2 j +2} |, D _L =| X _{2 i ,2 j} - X _{2 i ,2 j +2} |+| X _{2 i -1,2 j} - ₁ - X _{2 i -1,2 j +1} |+| X _{2 i +1,2 j -1} - X _{2 i +1,2 j +1} |, D _R =| X _{2 i ,2 j} - X _{2 i ,2 j +2} |+| X _{2 i -1,2 j +1} - X _{2 i -1,2 j +3} |+| X _{2 i +1,2 j +1} - X _{2 i +1,2 j + 3} |, where X _{2 i -1,2 j +1} , X _{2 i +1,2 j -1} , X _{2 i -1,2 j -1} , X _{2 i +1,2 j +1} , X _{2 i -1,2 j +3, X 2} i +1,2 j +3 lines the step (C), the calculated value of the diagonal of the blank pixels. The fast high-definition video image interpolation method according to claim 7, wherein the following formula is used in the step (E) to calculate the upward weight of the remaining blank pixels X _{2 i , 2 j +1} W _U , downward weight W _D , left weight W _L and right weight W _R : W _U =exp(- D _U / h ), W _D =exp(- D _D / h ), W _L =exp( - D _L / h ), W _R =exp(- D _R / h ), where h is the adjustment parameter. For example, the fast high-definition video image interpolation method described in claim 8 wherein the step (F) uses the following formula to calculate the horizontal direction of the remaining blank pixels X _{2 i , 2 j +1} Absolute pixel difference (Δ H ) and absolute pixel difference (Δ V ) in the vertical direction: The patentable scope of application of the nine fast the high quality video image interpolation method, wherein the step (G) in Comparative horizontal line direction of the absolute pixel differences (Δ H) and the vertical direction of the absolute pixel differences (Δ V) of The value determines the interpolation direction of the remaining blank pixels X _{2 i , 2 j +1} and its interpolation value. The fast high-definition video image interpolation method according to claim 10, wherein the step (H) further comprises the following steps: (H1) setting an initial high-resolution image I _H (0); (H2) Calculating an error e between the low-resolution image I _L and a simulated low-resolution image simulated by the high-resolution image I _H ( n ); (H3) according to the error e , for the high-resolution image I _H ( n ) performing correction and generating a corrected high-resolution image I _H ( n +1); and (H4) repeating steps (H1) through (H3) until a predetermined number of operations. A fast high-definition video image interpolation device receives a low-resolution image and inserts a pixel into the low-resolution image to increase the resolution thereof, thereby generating a high-resolution image, and the low-resolution image is arranged Forming a plurality of pixels in a two-dimensional array, wherein the pixels X _{2 i , 2 j} , i , j are positive integers, the image interpolation device includes: a video image interpolation processor that receives the low-resolution image and using a non-linear exponential function (exponential function), and a diagonal blank pixels X _{2 i +1,2 j +1} on adjacent left, upper right, lower left and lower right absolute four directions of the pixel difference calculated _{i + 1,2 j + 1} and the weight values of neighboring pixels of the blank pixels obliquely X _2, and _{2 i +1,2 j +1} value adjacent to a pixel based on the weight of the inclined gap and the pixel X , the oblique blank pixels X _{2 i +1,2 j +1} as a weighted average interpolation; line buffer, coupled to the interpolation processor of the video image to the interpolation processor temporarily storing video images arising Data; a table device connected to the video image interpolation processor, according to the video shadow The interpolation processor output indicators, the use of table lookup to real time of the exponential function computation; and the latter image enhancement processor, coupled to the line buffer means, to the high-resolution images for the image enhancement processing.